Using DFR to determine dissipation factor temperature dependence

Using DFR to determine dissipation factor temperature dependence by Matz Ohlen and Peter Werelius, Megger With an aging power component population, today s electrical utility industry faces a tough challenge as failures and consequent repair and revenue loss may inflict major costs. Transformers have become one of the most mission critical components in the electrical grid. The need for reliable diagnostic methods drives the world s leading experts to evaluate new technologies that improve reliability and optimise the use of the power network. Modern technology and developments in signal acquisition and analysis techniques have provided new tools for transformer diagnostics. Of particular interest are dielectric response measurements where insulation properties of oil-paper systems can be investigated. Dielectric frequency response(dfr), was introduced more than a decade ago and has been thoroughly evaluated in a number of research projects and field tests with good results. DFR data in combination with mathematical modelling of the oil-paper insulation is proven as an excellent tool for moisture assessment. Since the modelling theory contains influence of temperature, DFR and modelling can be used to calculate the temperature dependence of the insulation system. The condition of the insulation is an essential component of the operational reliability of electrical power transformers, generators, cables and other high voltage equipment. Transformers with high moisture content cannot sustain higher loads without risk. Bushings and cables with high dissipation factor at high temperature can explode due to thermal runaway. It is also very important to identify good units in the aging population of equipment. Adding just a few operating years to the expected end-of-life for a transformer or cable means substantial cost savings. Traditional dissipation factor measurements The most common insulation diagnostic test involves measuring capacitance and power factor at 50/60 Hz. Most tests are done at 10 kv (or sometimes Typical power factor values @ 20 C New Old Warning-alert limit Power transformers, oil insulated 0,2 0,4% 0,3,5% > 0,5% Bushings 0,2 0,3% 0,3 0,5% > 0,5% lower, depending on the voltage rating of the component), and at operating temperature, but there are also tests with variable voltage (tip-up/step-up testing) as well as tests where power factor versus temperature is measured. Analysis is based on (historical) statistics and comparison with factory values. Since insulation properties depend on temperature, temperature compensation has to be used for measurements not performed at 20 C. This is normally achieved by using temperature correction table values for certain classes of devices [1]. In IEEE 62-1995, typical power factor measurement values for transformers and bushings are categorised. Typical temperature corrections are shown in Fig. 1. It is obvious that the given values are approximate guidelines only. IEEE 62-1995 states; The power factors recorded for routine overall tests on older apparatus provide information on the general condition of the ground and inter-winding insulation of transformers and reactors. They also provide a valuable index of dryness, and are helpful in detecting undesirable operating conditions and failure hazards resulting from moisture, carbonisation of insulation, defective bushings, contamination of oil by Table 1: Typical power factor values (IEEE). dissolved materials or conducting particles, improperly grounded or ungrounded cores, etc. While the power factors for older transformers will also be <0,5% (20 C, power factors between 0,5% and 1,0% (20 C) may be acceptable; however, power factors >1,0% (20 C) should be investigated. Dielectric frequency response measurements The first field instrument for DFR measurements was introduced 1995 [2]. Since then the numerous developments of the technology have taken place and several international projects/reports define dielectric response measurements together with insulation modelling as the preferred method for measuring moisture content of the cellulose insulation in power transformers [3,4,5]. In DFR tests, capacitance and dissipation/power factor is measured. The measurement principle and setup is very similar to traditional 50/60 Hz testing with the difference that a lower measurement voltage is used (200 V) and instead of measuring at line frequency 50/60 Hz, insulation properties are measured over a frequency range, typically from 1 khz to 1 mhz. The results are normally presented as capacitance and/or tan delta/power factor versus frequency. The measurement setup is Fig. 1: Typical power factor temperature corrections. Fig. 2: DFR/FDS test setup. energize - April 2012 - Page 38

Fig. 3: DFR measurements on four different transformers at different temperatures with moisture content ranging from 0,3 to 3,4%. Fig. 5: MODS moisture analysis. Fig. 4: Parameters that effects the dissipation factor at various frequencies. Fig. 6: MODS analysis for two transformers with different oil quality and moisture content. shown in Fig. 2 and typical DFR results from measurement on transformers in different conditions in Fig. 3. Moisture assessment The capability of DFR to measure dissipation factor as function of frequency, gives the user a powerful tool for diagnostic testing. Moisture assessment is one example. High moisture levels in transformers is a serious issue since it limits the maximum loading capacity (IEEE Std C57.91 1995) and the aging process is accelerated. Accurate knowledge of the actual moisture content is necessary in order to make decisions on corrective actions, replacement/ scrapping or relocation to a different site in the network with reduced loading. Using DFR for determining moisture content in the oil-paper insulation inside an oilimmersed power transformer has been described in detail in several papers and articles elsewhere [3,4,5], and is only briefly summarised in this paper. The dissipation factor for an oil/cellulose insulation plotted against frequency shows a typical inverted S-shaped curve. With increasing temperature the curve shifts towards higher frequencies. Moisture influences mainly the low and the high frequency areas. The middle section of the curve with the steep gradient reflects oil conductivity. Fig. 4 describes parameter influence on the master curve. Using DFR for moisture determination is based on a comparison of the transformers dielectric response to a modelled dielectric response (master curve). A matching algorithm rearranges the modelled dielectric response and delivers a new curve that reflects the measured transformer. The moisture content along with the oil conductivity for the master curve is presented. Only the insulation temperature (top oil temperature and/or winding temperature) needs to be entered as a fixed parameter. Two different transformers are shown in Fig. 6. The two units have the same 0,7%, 50/60 Hz dissipation factor, characterised by IEEE 6 1995 as warning/alert status calling for investigation. The investigation is done as moisture analysis using MODS. The two transformers are very different and maintenance measures for the two would also be different. Transformer 1 has good oil but needs drying. Transformer 2 has low moisture but needs oil change or regeneration. Bushing diagnostics Aging/deterioration of high-voltage bushings is a growing problem and manufacturers as well as utilities and test system providers are suggesting and testing various methods for detecting bushing problems before they turn into catastrophic failures. This includes on-line monitoring as well as off-line diagnostic measurements [6,7]. Traditional 50/60 Hz dissipation/power factor testing may give an indication of aging/high moisture content, especially if performed at various temperatures as shown in Fig. 7, [8] and Fig. 8, [10]. As seen in Fig. 7: Dissipation factor (%) vs temperature for OIP bushings with various moisture content [6]. Fig. 8: Power factor (%) vs temperature ( C) for good and bad bushings, [10]. energize - April 2012 - Page 39

Fig. 9: Power factor at 60 Hz for oil impregnated cellulose insulation with various moisture contents as a function of temperature (ºC). Fig. 11: Bushing dissipation factor as a function of temperature. Measured and converted data compared to published data, [6]. Fig. 10: Relationship between power factor values at different frequencies taken at different temperatures. Fig. 12: Power factor as function of temperature (ºC) for four different transformers [11]. Fig. 7, dissipation factor values at lower temperatures are quite similar from very low to moderate moisture content. A significant change is not seen until measuring at about 50 C. The bad bushing in Fig. 8 can be compared with bushing data in Fig. 7. Estimated moisture content is about 4%. Increased dissipation factor at higher temperatures is a good indicator of bushing problems. Catastrophic bushing failures (explosions) are often caused by what is called thermal runaway. A high dissipation factor at higher temperatures result in an increased heating of the bushing which in turn increases the losses causing additional heating which increases the losses even further and the bushing finally explodes. Individual temperature correction (ITC) DFR measurements and analysis together with modelling of the insulation system includes temperature dependence. A new methodology (patent pending) is to perform DFR measurements and convert the results to dissipation factor at 50/60 Hz as a function of temperature. This technique has major advantages in measurement simplicity. Instead of time consuming heating/cooling of the bushing and doing several measurements at various temperatures, one DFR measurement is performed and the results are converted to 50/60 Hz tan delta values as a function of temperature. A result of the technique is shown and compared with the classical results presented by Blodget [9] in Fig. 9. The method is based on the fact that a certain power factor measurement at a certain frequency and temperature corresponds to a measurement made at a different temperature at a different frequency. The conversion calculations are based on Arrhenius' law/equation, describing how the insulation properties are changing over temperature. The relationship is depicted for three different activation energies in Fig. 10. Applying this technique on real-world DFR measurements on bushings gives results as shown in Fig. 11. Two bushings, OK and bad are compared with manufacturer's values from Fig 7, [6]. The bad bushing is estimated to have about 4% moisture and should be considered at risk. Temperature correction tables such as in IEEE/C57.12.90 give average values assuming average conditions and are not correct for an individual transformer or bushing. This was confirmed in field experiments and some utilities try to avoid applying temperature correction by recommending performing measurements within a narrow temperature range [11]. Examples are shown in Figs. 12 and 13. Power factor was measured at 10 kv on four transformers and three bushings of different age, condition and at various temperatures. Temperature dependence is very different for the transformers and bushings and using standard temperature correction tables will not give correct values for the 20 C reference value. With DFR and the technique for converting data to temperature dependence, it Fig. 13: Power factor as function of temperature (ºC) for three different bushings [11]. Fig. 14: Standard temperature correction compared with individual temperature correction for samples of GE Type U bushings. energize - April 2012 - Page 40

Fig. 15: Temperature correction for transformers in various conditions. Fig. 17: Tan delta at 50 Hz for dry Kraft paper as function of temperature. Fig. 16: Dissipation factor as function of frequency for dry Kraft paper. Fig. 18: DFR measurements and moisture analysis results at different temperatures. is possible to do accurate, individual temperature compensation. For a good component, the temperature dependence is weak. When the component gets older and/or deteriorated, the temperature correction factor becomes much larger, i.e. the temperature correction is a function of aging status. This is in line with several projects and studies [8,10]. Examples of individual temperature correction for bushings are shown in Fig. 14. Manufacturer's table data is only valid for as-new bushings. As soon as the bushing starts to show deterioration, the temperature dependence increases. Bad bushings have a very large temperature correction. Individual temperature correction for transformers is more complex compared to single-material components e.g. bushings. The oil-paper insulation activation energy constant W a in Arrhenius' law, for oil impregnated paper is typically 0,9 1 ev, while for transformer oil W a is typically around 0,4 0,5 ev Individual temperature corrections for transformers of various ages are shown in Fig. 15. transformer data is summarised in Table 2. As seen in the figure, each transformer has its individual temperature correction. New units have a negative correction for slightly elevated temperatures and will show dramatically different results if the standard table are used. Aged transformers show the same behaviour as the standard tables but with a much stronger temperature dependence compared to the average IEEE values. Experimental results Oil impregnated Kraft paper Samples of Kraft paper with various moisture contents was measured at different temperatures [13]. Results for dry paper, moisture content <0,5% is shown in Fig. 16 Using DFR technique to estimate temperature dependence based on measurements at one temperature only, gives the results shown in Fig. 17. As seen in the diagram, the calculated temperature dependence matches very closely to the actually measured dissipation factors. Transformers DFR measurements on a distribution Manufacturer Year Moisture Power rating Status Pauwels 2005 0,4% 80 MVA New, at factory Pauwels 2000 0,3% 20 MVA New, at utility Westinghouse 1985 1,5% 40 MVA Used, spare at utility Yorkshire 1977 4,5% 10 MVA Used and scrapped Table 2: Transformer data. transformer at various temperatures are shown in Fig. 18. As expected the moisture analysis (moisture in paper insulation) show the same values independent of insulation temperature (insulation temperature was estimated as winding temperature, measured as winding resistance). Oil and paper insulation must be treated separately when modeling a transformer to estimate temperature dependence. This is described in Fig. 18. Combining the modeling results and converting to temperature dependence gives the temperature curves in Fig. 20. Also for this insulation system containing two different temperature dependent materials, the conversion gives results very close to the actual measured tan delta values. Bushings An Asea/ABB GOB OIP bushing, used but expected to be in good condition, has been measured at different temperatures. Tan Delta and DFR measurements were performed at three temperatures; Indoor at 22 C, outdoor at -8 C and in a heated chamber at 42 C. Results are shown in Table 3. Calculated temperature corrections using DFR results are presented together with the manufacturer's average temperature correction data in Fig. 20. For the specific bushing, individual temperature correction (ITC) both at 22 C and 42 C fits very well with manufacturer's data, indicating a bushing in normal condition. energize - April 2012 - Page 42

Fig. 19: Dissipation factor as function of frequency for oil and cellulose insulation. Fig. 21: Temperature correction curves for ABB/ASEA GOB bushing. Insulation Temperature, C Measured Power Factor @ 200 V @ 1-10 kv Comments -8 0,86 1,04 1,14 22 0,46 0,46 42 0,34 0,32 Voltage dependent at low temperatures Table 3: Power factor measurents on ABB GOB bushing. Fig. 20: Tan delta at 50 Hz for a distribution transformer as function of temperature. Discussion The temperature dependence of the dissipation factor of an insulating material needs to be considered when comparing measurement results with previous tests or factory values. Historically this has been done by the use of average temperature correction tables. Results are disappointing and many asset owners try instead to perform diagnostic measurements at a specific (narrow) temperature range. The new method of using frequency data and calculate/model the temperature dependence of the actual component offers an alternative to waiting for the correct temperature and then do the test. It gives the possibility to have correct 20 C reference values and also to make a correct comparison to previously measured non-corrected data at other insulation temperatures. How accurate the individual temperature correction can be is a valid question. As presented in this paper, using standard tables can easily give power factor errors in the order of ±50 100% or more. The ITC examples presented show good correlation between the calculated and actually measured dissipation factor at various temperatures. However, envisioning a standard method used for a large population of components of various design and makes, a certain variation is anticipated. Preliminary tests with commercial test instruments and SW indicates that the inaccuracy for ITC is about ±5 10% at the extremes of a 5 50 C temperature range i.e. correcting from 5 or 50 to 20 reference. Summary and conclusions Dielectric Frequency Response (DFR/FDS) measurement is a technique/methodology for general insulation testing and diagnostics. In comparison with standard 50/60 Hz dissipation factor measurements, DFR measurements provide the following advantages: Capability of performing individual temperature correction of measured 50/60 Hz dissipation/power factor at various temperatures to values at reference temperature, 20 C. Capability of comparing test results from a new measurement at a certain temperature to another measurement at a different temperature Capability of estimating dissipation/ power factor at operating temperature in order to assess risk of thermal runaway catastrophic failure. Capability of estimating the moisture content of oil-immersed cellulose insulation in power transformers and bushings Capability of investigating increased dissipation factor in power components The insulation properties are very important for determining the condition of a power system component. Knowing the condition helps to avoid potential catastrophic failure and identifying good units and decide upon correct maintenance can save significant money due to postponed investment costs. References [1] "IEEE Guide for Diagnostic Field Testing of Electric Power Apparatus; Part 1: Oil Filled Power Transformers, Regulators, and Reactors, IEEE 62-1995. [2] P Werelius et al, Diagnosis of Medium Voltage XLPE Cables by High Voltage Dielectric Spectroscopy, paper presented at ICSD 1998. [3 ]U Gäfvert et al, Dielectric Spectroscopy in Time and Frequency Domain Applied to Diagnostics of Power Transformers, 6th International Conference on Properties and Applications of Dielectric Materials, June 21-26, 2000, Xi'an, China. [4] S M Gubanski et al, "Dielectric Response Methods for Diagnostics of Power Transformers, Electra, No. 202, June 2002, pp 23-34 also in CIGRE Technical Brochure, No. 254, Paris 2004. [5] S M Gubanski et al, Reliable Diagnostics of HV Transformer Insulation for Safety Assurance of Power Transmission System. REDIATOOL - a European Research Project, paper D1-207 CIGRÉ 2006. [6] Swedish Bushings Plant Sees Growth in RIP Designs, INMR Quarterly, Issue 68, 2005. [7] J M Braun et al. Accelerated Aging and Diagnostic Testing of 115 kv Type U Bushings, paper presented at IEEE Anaheim 2000. [8] C Kane, Bushing, PD and Winding Distortion Monitoring, paper presented at ABB Seminar Power Transformer Health Monitoring and Maintenance Johannesburg 2008. [9] R B Blodget, Influence of Absorbed Water and Temperature on Tan Delta and Dielectric Constant of Oil-Impregnated Paper Insulation, Trans. AIEE, 1961. [10] R Brusetti, Experience with On-line Diagnostics for Bushings and Current Transformers, NETA Fall 2002, paper A335. [11] R K Tyagi, S Victor, N S Sodha, Application of Temperature Correction Factors for dissipation factor Measurements for Power Transformers A case study, Doble Client Conference, Vadodara, India 2006. [12] P Werelius, M Ohlen, Dielectric Frequency Response Measurements on Power Transformers, EuroTechCon 2008, Liverpool, UK. [13] R Niemanis et al, Determination of Moisture Content in Mass Impregnated Cable Insulation Using Low Frequency Dielectric Spectroscopy, IEEE Power Engineering Society Summer Meeting 2000, Seattle, Washington, USA. Contact Marius Pitzer, Megger, Tel 021 557-6572, marius.pitzer@megger.com energize - April 2012 - Page 43